RETRACTED: Proteasome storage granules protect proteasomes from autophagic degradation upon carbon starvation

Essential revisions:

A key finding of this work is that formation of PSGs correlates with more rapid recovery from starvation. Most of the data on this point are correlative, and it is difficult to rule out that the treatments used do not exert their effect through perturbing or promoting some process other than PSG formation. With this in mind, there are two experiments that would strengthen this key section of the paper that the reviewers believe should be done:

1) Repeat the ∆ubp3 rescue experiment with a mutant other than ∆blm10 (e.g. ∆nat3; a ∆nat3 ∆ubp3 strain should recover from carbon starvation faster than ∆nat3), to establish that this effect is not limited to one particular mutant/treatment.

We agree that much of the data regarding PSG formation and its relation to cellular fitness is indeed correlative, but we feel that our conclusions are strongly based on the fact that so many different mutations and treatments led to results that pointed in the same direction. As such, we agree with the reviewers that performing the additional Δubp3 rescue experiments would be important. To address this point we undertook two approaches. The first involved assessing the effect of combining the Δnat3 and Δubp3 mutations on i) the localisation of Pre10-GFP and Rpn5-GFP as determined by confocal microscopy, and ii) the release of free GFP as determined by immunoblot. These data (now included in a new Figure 7—figure supplement 1A, B, C and D) clearly demonstrate that, while Pre10-GFP is delivered to the vacuole upon carbon starvation in Δnat3 cells (which do not form PSGs), this delivery is blocked in Δnat3 Δubp3 double mutant cells. Likewise, while autophagic release of free GFP from the Pre10-GFP reporter is readily detectable by immunoblot in Δnat3 cells following carbon starvation, this release is greatly reduced in the Δnat3 Δubp3 double mutant. These results now confirm that Ubp3 is required for delivery of the CP to the vacuole in two separate mutant strains; Δblm10 (Figure 7A and C) and Δnat3 (this new data). Interestingly, delivery of Rpn5-GFP to the vacuole, and release of free GFP from this reporter, was not blocked in the Δnat3 Δubp3 mutant, confirming the finding of both ourselves and Waite et al. (2016) that Ubp3 activity is required only for delivery of the CP to the vacuole upon starvation.

The second approach involved performing the yeast growth assays to monitor cell fitness following starvation with the Δnat3 single mutant and the Δnat3 Δubp3 double mutant, as suggested by the reviewers. These data (included in a new Figure 8K, L and M (replacing the original panels)) again correlate well with the data obtained for the Δblm10 and Δblm10 Δubp3 mutants. While the Δnat3 and Δnat3 Δubp3 cells grow slightly more slowly than wild type in the absence of starvation, growth recovery of the Δnat3 strain following carbon starvation is substantially reduced compared to the control (taking approximately 3 to 4 hours, compared to less than 1 hour for wild type). However, when the Δnat3 and Δubp3 mutations were combined, the cells were better able to resume growth following restoration of the carbon source than were Δnat3 cells alone (Figure 8K, L and M). These results add additional support to our conclusion that the extent of proteasome degradation upon carbon starvation impacts the ability of cells to recover from the starvation treatment.

2) Examine the effect of ∆ubp3 on recovery from starvation in ∆spg5. This is a key experiment because proteaphagy in carbon-starved ∆spg5 is not blocked by introducing ∆ubp3. Thus, a ∆spg5∆ubp3 strain should not show more rapid resumption of growth compared to ∆spg5, in the assay shown in Figure 8K. This would be particularly striking because ∆ubp3 does have a beneficial effect when this assay is performed with ∆blm10, and thus would greatly strengthen the correlation between low proteaphagy and rapid resumption of growth.

We again agree with the editors and reviewers that this experiment would provide strong additional support for our model. As such, we performed the experiment, and the data is now included in a new Figure 8N, O and P (replacing the original panels K, L and M). As anticipated by the reviewers, while the Δspg5 mutant cells showed a modest reduction in growth recovery following starvation (the effect is less severe than for the Δblm10 and Δnat3 mutants, as deleting Spg5 merely delays, rather than completely blocking, the delivery of the RP into PSGs; it may also reflect the fact that maintaining a pool of the CP is more important to growth resumption than is preserving the RP), combing the Δspg5 and Δubp3 mutations did not lead to faster growth recovery, in contrast to the effects seen when Δubp3 was combined with Δblm10 or Δnat3. Both this data and the new data from point 1 now considerably strengthen the observed correlation between low levels of proteaphagy and rapid resumption of cell growth following renewed availability of nutrients.

In addition, given that this paper shows for the first time the behavior of PSGs in Arabidopsis, the authors should evaluate whether these structures are reversible as they are in yeast.

We thank the reviewers for this suggestion. We subjected 5 day-old PAG1:PAG1-GFP pag1-1 seedlings grown in the light in sucrose-containing medium to 24 hours of fixed-carbon starvation by switching them to sucrose-free medium and growing them in the dark to prevent photosynthesis. We then returned the seedlings to a sucrose-containing medium and re-exposed them to the light, and monitored the fate of the cytoplasmic PSG-like structures over time. These results are now presented in a new Figure 9F. As can clearly be seen, the strong nuclear and diffuse cytoplasmic fluorescence signal observed in root cells grown in nutrient-rich medium was greatly reduced, and was instead found in bright cytoplasmic puncta when the seedlings were subjected to fixed carbon starvation. However, when the carbon source was replenished, these puncta began to disperse, and the cytoplasmic and nuclear signal returned. This was visible even 1 hour following the cessation of starvation, and after 4 hours the cytoplasmic puncta had essentially disappeared. The rapid reversibility of these punctate structures in Arabidopsis lends further support to our proposal that these are PSGs, and that they perform an equivalent function in plants as they do in yeast.

Reviewer #1:

The authors have examined the ability of cells of different genotype or treated with different agents to recover upon refeeding following nitrogen, carbon, or nitrogen plus carbon starvation. These experiments paint a consistent picture: if a cell can form PSGs upon starvation to protect its proteasomes from proteaphagy, it recovers more rapidly upon refeeding. My one comment here is that I would like to see the ∆ubp3 rescue in more than one example. Currently, the authors show that ∆ubp3 will rescue the recovery defect of a ∆blm10 mutant. Is that same true for ∆nat3, for example?

We thank the reviewer for their suggestion, and refer them to point 1 of the essential revisions described above.

The other major concern I have, which is that it is hard to exclude that PSGs form for some other reason other than to protect the proteasome from proteaphagy. The authors could address this by being a bit more circumspect in their Discussion. The recovery experiments described in the prior paragraph provide further support for this contention, but given that all of the mutants analyzed are likely to impact multiple physiological processes, it is difficult to exclude that the major protective effect of Blm10 or Nat3 is through some process other than PSG formation.

We apologise to the reviewer for not addressing their second major concern, but unfortunately this was not communicated to us following the initial review of our pre-submission. We agree with the reviewer that, on their own, each mutant or treatment condition we have studied (the Δblm10, Δnat3, Δspg5, Δubp3, rpn11-m1 and rpn11-m5 mutants, plus growth at high or low pH and treatment with 2-DG) likely impacts multiple cellular processes, and thus we cannot exclude that their major effects on proteaphagy, and the positive effect some of them impart on cell growth recovery after starvation, are mediated through other processes. However, we do feel that the overall weight of the evidence, with multiple mutants and treatment conditions all showing results consistent with our hypothesis, favours our conclusion that PSG formation protects proteasomes from autophagic degradation and helps improve cell fitness following recovery from starvation. We have taken the reviewer’s advice and modified our discussion to take into account the very valid point that was raised and added the sentence below to the Discussion:

“While we cannot exclude the remote possibility that PSGs also have alternative functions, and/or that the ability of the various factors studied here to protect proteasomes from autophagy arises from processes unrelated to PSGs, the sum of our results strongly converges to this conclusion.”

Another issue that is not addressed is why formation of PSGs would protect from autophagy? There is a paragraph in the Discussion devoted to this topic but it focuses almost entirely on how/why PSGs form and says little about how formation of these structures would prevent autophagy. Autophagy can clearly process large structures including mitochondria and IPODs. Why would formation of PSGs be protective? Perhaps the authors can include a few lines on this in the Discussion.

We again agree with the reviewer that this is an interesting point worthy of inclusion in our Discussion, and we have thus modified it accordingly (see Discussion, ninth paragraph). We are currently lacking experimental data addressing how exactly PSGs (and potentially other biomolecular condensates) are able to evade the autophagy machinery, and so the newly added text is necessarily rather speculative. However, data in our new Figure 1—figure supplement 2C obtained in response to point 1 raised by reviewer 2 indicates that PSGs can form independently to the PAS, and there is thus at least a degree of spatial separation between PSGs and the autophagy machinery (this has been mentioned in the aforementioned paragraph). Whether there is some particular biophysical or structural property of PSGs (and maybe other condensates) that allows them to evade autophagy is an interesting question that to our knowledge has not yet been addressed in the literature.

Reviewer #2:

[…] Overall, this study is well performed and presented. The methods used in this study are appropriate. Remaining questions would be how PSGs are able to escape from autophagic degradation and how de-ubiquitination is important for proteaphagy. However, these could be future studies as the authors have made a significant progress in revealing a role of PSG formation in protection from autophagic degradation and cell regrowth after starvation.

We thank the reviewer for their positive comments about or work. We agree with reviewer #2 that how PSGs escape capture by the autophagy machinery, and precisely how de-ubiquitylation fits into the picture, are important question that we intend to address in the future.

Reviewer #3:

[…] I think the paper would be suitable for publication in eLife after the modest criticisms, suggestions, and corrections below are addressed:

o The apparent mass of the cleaved GFP in the western blots from yeast extracts (Ex. Figures 1A-1C) was larger than what would be expected. In contrast, the free GFP MW from plant protein extracts was close to the expected size (Figures 9A, 10A). Please explain this difference.

As the reviewer correctly points out, there is a discrepancy in the size of the free GFP observed between yeast and plant cell extracts on our immunoblots. We have noticed this before, and indeed it has been a consistent feature of the experiments we have performed with yeast that the free GFP band runs at around the 30-31 kDa mark (see e.g. Marshall et al., 2016), rather than around 27-28 kDa in the plant cell extracts. Unfortunately, we currently have no satisfactory explanation for this phenomenon. One possibility is that plant and yeast vacuolar hydrolases may have different cleavage specificities, meaning that the GFP gets processed at a different position. Alternatively, there may be some differences in buffer composition between our yeast and plant cell extraction methods that cause this difference. A brief review of the autophagy literature does reveal considerable discrepancy between the size of the free GFP band between (and even within) different studies, ranging from smaller than 20 kDa (e.g. in Lu et al., 2014) up to 30-31 kDa (e.g. in Müller et al., 2015), but the reasons for this are not yet clear.

o Can the cytoplasmic foci (Figure 1D) be rapidly remobilized to the nucleus after switching to fresh YPGA that has 2% glycerol? Similarly, can the 2-DG-induced cytoplasmic foci (Figure 3E) remobilize upon switching to fresh +N +C medium?

We thank the reviewer for this suggestion, and have performed the suggested experiments, which are now included in new Figure 1—figure supplements 2D and 2E. In either case, the cytoplasmic foci observed upon carbon starvation or 2-DG treatment were rapidly re-mobilised upon switching to fresh carbon-containing medium, or medium lacking 2-DG, respectively. This occurred within 30 minutes following cessation of starvation (the earliest time point examined), and within 1 hour of the removal of 2-DG (the slightly slower resorption rate following 2-DG treatment is likely due to incomplete removal of the chemical following wash-out). This timing is consistent with the previous study by Laporte et al. (2008), who showed that PSGs are rapidly re-mobilised within 15 minutes following exit from quiescence.

o What percentage of cells have separate cytoplasmic foci of Pre10-GFP and Rnq1-mCherry in Figure 1H?

To address this question we grew and imaged more cells expressing PRE10-GFP and RNQ1-mCherry, and quantified the number of cells containing distinct GFP and mCherry puncta following carbon starvation. Over 90% of cells (from over 200 quantified) had these distinct puncta, and this is now mentioned in the fifth paragraph of the subsection “Proteasomes are rapidly degraded by autophagy in response to nitrogen but not carbon starvation”.

o As shown in Figure 5F, some mCherry-Spg5 puncta colocalize with Rpn5-GFP in carbon-starved cells, which seems inconsistent with the description in the last paragraph of the subsection “Spg5 helps deliver the RP to PSGs and protects the RP from proteaphagy”. Similarly, some mCherry-Spg5 puncta colocalized with Pre10-GFP (Figure 4—figure supplement 1C); again this is inconsistent with the description in text. Please address this apparent discrepancy, and more importantly, provide the percentage of PSGs marked by Rpn5 or Pre10 that co-localize or not with Spg5.

We thank the reviewer for noticing that our description of the co-localization data was not as precise as it should have been. They are indeed correct that there are rare instances in which mCherry-Spg5 does indeed appear to co-localize in cytoplasmic puncta with either Pre10-GFP or Rpn5-GFP. As such, we have modified the Discussion in the last paragraph of the subsection “Spg5 helps deliver the RP to PSGs and protects the RP from proteaphagy”, to make this clear. In addition, we grew and imaged more cells expressing either PRE10-GFP or RPN5-GFP together with mCherry-SPG5, and quantified the number of cells containing overlapping GFP and mCherry puncta following carbon starvation. We found that 12% of cells (from over 200 quantified) had puncta containing both Rpn5-GFP and mCherry-Spg5, while the remaining 88% showed puncta containing Rpn5-GFP only. Similarly, just 6% of cells (from over 200 quantified) had puncta containing both Pre10-GFP and mCherry-Spg5, while the remaining 94% showed puncta containing Pre10-GFP only. This quantification has been included in the aforementioned paragraph.

o In Figure 7C, what is the GFP cleavage profile for Pre10-GFP and Rpn5-GFP in ∆ubp3 mutant?

We thank the reviewer for pointing out this omission from our original experimental design. To address this, we have replaced the immunoblots in Figure 7C with new blots that include the GFP cleavage profile of Pre10-GFP or Rpn5-GFP in Δubp3 cells alone, together with the other genotypes included in the original Figure. In both cases, the absence of Ubp3 alone did not cause additional release of free GFP compared to wild-type cells.

o For the starvation treatments on 5-d-old Arabidopsis seedlings, is there data indicating the nutrients from seeds are exhausted at this time point?

We thank the reviewer for raising this interesting point. There is indeed literature describing the mobilisation of seed storage reserves upon seed germination in Arabidopsis. The major nitrogen sources are the seed storage proteins, which in Arabidopsis are 12S globulins and 2S albumins. These are known to be fully degraded within 4 days of seed germination (see, for example, Figure 6 of Hunter et al., 2007 and Figure S3A of Gao et al., 2015). Likewise, the major carbon reserve in Arabidopsis seeds is considered to be triacylglycerol (TAG), and the TAG marker fatty acid eicosenoic acid (20:1) was shown to be more than 95% degraded by day 5 following germination (see Figure 2A of Penfield et al., 2005). As such, we feel that the day 5 time-point selected by us for the Arabidopsis starvation experiments is appropriate. This information, along with the relevant citations, has now been included in the Results subsection “PSG assembly and the protection of proteasomes from proteaphagy are conserved in Arabidopsis”.

Also, ConA was only added to the medium for starvation treatments. Could ConA cause proteasome relocalization to form cytoplasmic foci in Arabidopsis even when grow in nutrient rich medium?

We thank the reviewer for noticing this issue; they are indeed correct that our original experimental design did not preclude the fact that the cytoplasmic puncta observed upon fixed carbon starvation could be caused by the ConA treatment. To address this, we now include data in a new Figure 10—figure supplement 1 confirming that the PSG like structures still form upon fixed carbon starvation even in the absence of ConA treatment. This experiment has the additional advantage of confirming that these PSG-like puncta are likely resident in the cytosol, rather than being stabilised by the ConA treatment inside the vacuole.

o In Figure 9, can these PSG-like structures in Arabidopsis be remobilized into nuclei upon carbon restoration? It would be nice, but not essential, in establishing the parallels with yeast PSGs, to show whether these PSG-like structures in Arabidopsis roots behave similarly compared to the PSGs in yeast.

We thank the reviewer for raising this point, and agree with them that inclusion of this data strengthens our conclusion that the punctate structures containing PAG1-GFP and RPN5a-GFP that are observed in Arabidopsis roots upon fixed carbon starvation are indeed similar to PSGs. Please see point 3 in the essential revisions for details of how we addressed this point.

o There is no description of the methods used for plant protein extraction as far as I can tell.

We thank the reviewer for noticing this omission. A description of the method used for plant protein extraction has now been included in the Materials and Methods subsection “Immunological techniques”.

o I did not see any details or citation for the atg7-2 mutant line used in Figure 10D.

We once again thank the reviewer for noticing this omission. The relevant accession number for this line (GABI_655_B06), and the reference in which it was first described (Chung et al., 2010), are now included in the Materials and Methods subsection “Arabidopsis materials and growth conditions”.